Abstract:

A reconstruction method which does not depend on pure numerical data was presented. Each sample of the dataset is corresponding to a group of genes, which is contained in one pathway branch. Under the premise that genes represent indifferent samples may share the same transduction order, a firstorder Markov chain model is built; then an ExpectationMaximization (EM) algorithm is used to estimate model parameters so as to reveal the regulative among genes; finally, the network topology is reconstructed according to the above results. The proposed method can compensate the insufficiency of many reconstruction algorithms and has more biological significance. The effectiveness of the algorithm is validated by reconstruction of Protein Kinase A (PKA) and MitogenActivated Protein Kinase (MAPK)/Extracellular signalregulated Kinase (ERK) pathways.

Key words: artificial intelligence, network reconstruction, signaling pathway, markov chain, EM algorithm